Transfer-Based Deep Learning Technique for PCOS Detection Using Ultrasound Images
Narinder Kaur, Ganesh Gupta, Paramjot Kaur
Abstract
Polycystic ovary syndrome, or PCOS, is a complex problem of hormones that affects up to 10% of women of childbearing age. Correct treatment and control of PCOS depend on a correct diagnosis and classification of the disease. The majority of people who have PCOS are overweight, have more severe symptoms, have irregular periods, and have an increased amount of body hair. It is critical to acquire a diagnosis of polycystic ovary syndrome (PCOS) as quickly as possible so that you may begin treatment for the symptoms and reduce the associated health concerns. At the moment, physicians and radiologists employ ovarian ultrasonography to detect PCOD by determining the size of the eggs and calculating the number of eggs present in the ovary. This is one of the PCOS diagnosis variables that might be challenging. In this work, we suggest a way to classify people with PCOS by using ultrasound images of the ovaries and transfer learning-based deep learning models. Our model is trained on a large set of ultrasound images and was already trained using predefined InceptionV3 models on a set of medical images in general. We were able to classify PCOS with 99.48% accuracy, which shows that transfer learning is a good way to improve the accuracy of PCOS detection. Our experimental results demonstrate that transfer learning-based classification achieves superior performance in PCOS subtype classification compared to traditional machine learning approaches. Clinicians may be able to better diagnose and treat PCOS with the help of our suggested model.